Actions

Module System

From Gambit wiki

Revision as of 21:57, 30 March 2008 by Pflanze (talk | contribs) (major overhaul)

This is a project to create a module system for Scheme. It does not have a real name yet. For now, it will target the Gambit Scheme system.

The system is aiming to be complementary to the Snow package system: unlike Snow, this project does not focus on package distribution, but instead it has the following objectives:

  • good interactive development capabilities (runtime inspection and modification, included light-weight build system for the base case to make running code as easy as in any scripting language)
  • good code modularity features (parametrization, optional hygienic macros, infrastructure for generics, ..)
  • generality (run code written in different language styles, to some extent in different module philosophies, interface with object systems, host languages requiring translation like lazy-by-default variants of Scheme or maybe different languages like Arc or XSLT)
  • efficiency (employ cross-module inlining, parametrization specialization)
  • attempt portability by moving parts into or building parts on top of the Snow framework (possibly not achieving the same level of integration with the host system without implementing suggested changes of the latter, but at least make modules written by users portable to other Scheme systems)

The projects wants to deliver a base system usable for everyday use quickly, and explore the more advanced objectives in the longer term.

The primary long-term aim is to be very open for extension, by offering open access to the modularization mechanisms and finding the right mechanisms, so as to make it possible to integrate code from different language and modularization philosophies as cleanly as possible, and to make it easier to contribute towards this goal. We are aware that this is a noble aim and that the development will involve many steps of learning and searching for solutions.

During this process, several systems may be developed which either help daily code development on Gambit or contribute to the understandings and code base for attaining the longer term goals.

Currently we have the following development steps in mind:

  1. during a first phase, the project will continue to develop within the code bases of the chjmodule and the gambit-modules projects.
  2. implement a better interface language (like e.g. that of Scheme-48) on the chjmodule code, targetting parametrization
  3. investigate ideas like those from the CL Lexicons project for making easier to write macros (without automatic hygiene)
  4. investigate what automatic hygiene systems exactly need, implement necessary infrastructure
  5. introduce hygienic macros, maybe by partial adaption of existing code bases (Abdulaziz Ghuloum's, psyntax, Riaxpander, Scheme-48's, Andre v. Tonder's)

Before, during or after steps 4-5 modules with a lazy-by-default variant of Scheme will be implemented, as an additional play ground for working on the right hooks for code transformations.

Leading Contributors

How to contribute

One important part of the task involves getting to know the Gambit internals well enough to be able to embed the system well. Partly, the current leading contributors can help with this, partly this has to be worked out (by reading code, asking Marc, and writing documentation).

Another essential one is laying a sound foundation especially for syntax handling (macros), considering the limited knowledge of the current contributors, this involves reading the right literature and discussing with more experienced people.

Seeking contact with people in other module development projects is an important duty; we're currently especially thinking of those working with R6RS implementations, and those working on module systems for Common Lisp (Lexicons, PLT-alike module system on CL GSoC proposal by LispNYC).

Google Summer of Code

This project is attempting to participate in the Google Summer of Code campaign under the hospice of LispNYC and offers to mentor a student. There are several possibilities for contribution, depending on the background of the student. One is to tackle one or more of the above development steps. Another one could be to work with the PLT-for-CL project and port it to our infrastructure.


Timeline

Rough points of the planned timeline:

  • James Long is reading literature, and experimenting with several existing Scheme module systems
  • Christian Jaeger finishes up and announces the long-promised new chjmodule release
  • seeking contact with the other module projects, student selection if possible
  • seeking more input, continuing to work on chjmodule and gambit-modules to proceed with the above development step list

Virtual Library

We are keeping a growing list of literature which we think is relevant for our work. It contains items we've read or intend to read, so it will be growing; we will not put anything there which we don't think we will manage to read, so by looking at that list you can see whether we are missing something you think we should know about. We hope that you will notify us in this case.

Log

There is a logfile mentioning the events in the project.

License

The project intends, for the time being, to license all code under the same licenses as Gambit, namely dual LGPL version 2.1 or Apache version 2.0 license (freely choosable). Requirements for making the code portable may lead us to change the license in due course.